應用區域化變數理論於短期風速估測

Short-Term Wind Speed Prediction Using Theory of Regionalized Variables

王茲元(Tzu-Yuan Wang)

指導教授 ： 楊宏澤

中原大學/工學院/電機工程研究所/碩士(2006年)

https://doi.org/10.6840/cycu200600646

摘要

風速的連續變化將導致風力發電機組的輸出功率隨之改變，因此風電能轉換系統(wind energy conversion system, WECS)無法如同傳統發電機組進行調派。因此，若是能夠廣泛且準確的預測風速變化，不但可以作為電力系統發電機組調度與排程的參考，降低運轉成本，更可以進一步作為日後電力交易市場中，預測電力價格的參考依據之ㄧ。然而，於各可能設置風機地點預先架設觀測站進行風速量測與紀錄，則不僅耗時費事，所費亦不貲。
　　本研究之主要目的在於提出一估測模式，透過多處已量測紀錄風速之觀測站，推估其餘未設觀測站地點之風速資料。方法為結合各觀測站本身時間序列關係式與利用區域化變數理論之克利金推估法所得區域中各觀測點空間關係的推估模式，文中分別逐一刻意假設新竹、新屋、竹東、竹南、南庄其中之一為未知觀測站，其餘為已知觀測站，並採用實際數據資料進行本文所提短期風速估測模式的驗證和測試。
研究結果顯示各觀測站之短期風速估測皆以一階自迴歸移動平均模式ARMA(1,1)為佳，新竹與竹南兩處觀測站因具有較高的克利金變異數，顯示其觀測資料較具有代表性，應自行設置風速觀測站，並在推估其餘無觀測站風速資料時需納入其量測資料以為參考。

關鍵字

時間序列分析 ； 區域化變數理論 ； 短期風速預測

並列摘要

The continuous fluctuation of wind speed makes the power generated by wind turbines change accordingly. As a result, the wind energy conversion systems (WECS) cannot be dispatched like the conventional generating units.
Therefore, the more accurately and extensively short-term wind speed can be forecasted, the more efficiently the power generation can be scheduled. The wind speed forecasting will help decrease cost of conventional power generation. Further, it can also be employed as a reference for prediction of the spot price of electric energy in the deregulated power market. However, it would not only take lots of time and efforts but also be costly to build a wind-speed measuring and recording (M&R) station at every possible location to install the WECS.
The goal of this study is to develop an estimation model to forecast the wind speed of a location without an M&R station through the data collected from locations with M&R station. The approach used in the thesis is to combine the time series models of the data for the M&R stations themselves and the spatial structure relationships between locations. The study assumes in turn purposely one of Hsin-Chu, Hsin-Wu, Chu-Ton, Chu-Nan, Nan-Chuang with M&R stations as the location without M&R station, and the rest are locations with M&R stations to test and verify the proposed estimation approach.
Simulation results show that the time series model of ARMA(1,1) is more appropriate than the others. Furthermore, Hisn-Chu and Chu-Nan are the locations where an M&R station should be installed, due to a high estimated Kriging variance. The data collected at these two locations are also vital and should be included to estimate the unknown wind speed at the location without an M&R station.

並列關鍵字

Short-term Wind Speed Forecasting ； Regionalized Variable Theory ； Time Series Analysis

參考文獻

[3] 德商英華威公司，台灣風力發電產業的發展障礙。

[32] 彭百君，直接空調負載控制曲線自動化分類系統，中原大學電機工程學系，碩士學位論文，民國93年7月。

[14] G. N. Kariniotakis, G. S. Stavrakakis, E. F. Nogaret, ”Wind Power Forecasting Using Advanced Neural Networks Models”, IEEE Transactions on Enegy Conversion, Vol. 11, No. 4, pp. 762-767, December 1996.

[15] Shuhui Li, D. C. Wunsch, E. A. O’Hair, M. G.　Giesselmann, ”Using Neural Networks to Estimate Wind Turbine Power Generation”, IEEE Transactions on Energy Conversion , Vol. 16, No. 3,pp. 276-282, September 2001.

[16] M. C. Alexiadis, P. S. Dokopoulos, H. S. Sahsamanoglou, ”Wind Speed and Power Forecasting based on Spatial Correlation Models”, IEEE Transactions on Energy Conversion, Vol. 14, No. 3,pp. 836-842, September 1999.